5.1. Identification of the IRAS counterpart
Due to poor telescope pointing of the ESO 3.6-m at the time of the observations, the counterparts of 3 IRAS sources (13529-5934, 15066-5532, and 17088-4221) could not be determined unambiguously, even after careful analysis of the images. Due to the poor spatial resolution of IRSPEC , the spectrum of IRAS 16130-4620 was useless due to blending. We won't consider these objects any further, so that 16 objects remain for further discussion.
Based on the Br spectra we can identify objects of three different types: those with a Br emission line, those with a Br absorption line, and those with no Br line at all. We detected Br in absorption in 7 out of 16 objects. The absorption lines are very narrow in 6 objects, indicating a low surface gravity. This is a strong indication for the post-AGB nature of these objects. Six objects show Br in emission. Two of these also show a photospheric absorption profile. All emission line sources have a strong underlying continuum, unlike normal PNe. In another three objects no clear Br absorption or emission was visible.
5.3. Radio continuum
As noted in Sect. 3.3, the predicted optically-thin radio flux values (assuming Case B recombination) appear to be at least a factor of ten higher than the observational flux limits for the post-AGB objects. This indicates that for these objects either the radio flux is optically thick at 3 cm, or the Case B assumption is not valid.
After a star leaves the AGB, its mass loss decreases by several orders of magnitude while simultaneously the velocity of the wind increases. Hence the star will be surrounded by an increasingly more tenuous wind inside a detached AGB shell. At first the star will be too cool to cause any significant ionization, either in the wind or in the AGB shell. However, as the stellar temperature increases, ionization will start and the ionization front will steadily move outwards. This will also be the case for the radius at which the 3 cm radiation becomes optically thick (for brevity we will call this the 3 cm radius, ). Optically thick radio emission from a sphere of radius and a temperature Te gives a flux density at frequency of
where D is the distance to the star.
Using this approximation we calculated that 3 cm, assuming 0.7 mJy, = K, and D = 1 kpc. This upper limit is similar to the size Knapp et al. (1995) determined for the post-AGB stars CRL 915 (0.30 mJy) and IRAS 17423-1755 (0.44 mJy) using the same assumptions. As long as the ionization front has not reached the AGB shell, the 3 cm radius will not change much, because the outer regions in the density profile contribute very little to the optical depth, and consequently the radio flux will remain very low. However, once the post-AGB star reaches a temperature hot enough to ionize the AGB shell, the 3 cm radius will quickly increase by roughly two orders of magnitude, causing a dramatic increase in radio flux. This marks the onset of the PN phase. As an example, the sizes of the young PNe CRL 618 and IRAS 21282+5050 measured by Knapp et al. (1995) are a factor 10 larger than the sizes of post-AGB star nebulae. Their radio flux values, 67 mJy and 4.3 mJy respectively, are well above our detection limit.
In our post-AGB star candidates with Br in emission, the ionized region where the emission originates must be very small and dense. Probably, the AGB shell of our objects is not yet ionized, but the post-AGB wind could be. The Br spectra are unresolved at a resolution of 150 km s-1. Hence the wind velocity couldn't be much larger than this value. Evidence for a wind emanating from some of these central stars was presented in Van de Steene et al. (2000).
5.4. Spectral energy distribution
One of the well-established characteristics of post-AGB stars is that their Spectral Energy Distributions (SEDs) have a `double-peaked' shape. The two peaks in the spectrum correspond to the stellar and dust emission components. Post-AGB stars have been classified into four classes based on the shape of the SED by van der Veen et al. (1989).
Note that CLASS IVa´ was not contained in the original definition, but was added to classify objects that did not fit in any of the original categories.
The SEDs of the objects are shown in Fig. 10. The objects have the typical post-AGB SEDs as cited above. The SED class of each object is listed in Table 5. Six of the 16 positively identified objects are of CLASS II and the other 10 of CLASS IV. For objects in CLASS II the circumstellar dust is so optically thick that almost all star light is absorbed by the dust and is re-radiated at mid- to far-infrared wavelengths. The large infrared excess is commonly attributed to the presence of a very compact circumstellar dust shell and/or ongoing mass loss which obscures the central star from view. Objects in CLASS IV have less obscured central stars: the thermal emission from their circumstellar shells appear as a peak in the far-infrared and the central stars show up as a peak in the near-infrared (IVa) or optical (IVb). We always see some stellar signature in the near-infrared and therefore have no objects of CLASS III. For instance, Van der Veen et al. (1989) classified IRAS 16594-4656 as CLASS III, while we classified it as IVa. We extended the definition of CLASS IVa to include objects which show a clear stellar signature in the near-infrared, but peak in the K -band, just beyond 2 µm. (e.g. IRAS 13428-6232, IRAS 16279-4757). In the table these objects are marked as CLASS IVa´. These objects have no optical counterpart in the USNO catalog. When objects in CLASS IV peak at shorter wavelengths they usually have an USNO counterpart. The central stars of objects in CLASS IVb are less reddened and brighter than objects in CLASS IVa and have bright optical counterparts (e.g. IRAS 14325-6428, IRAS 14488-5405).
We especially draw attention to two objects which have unique SEDs. IRAS 15544-5332 is the only object in the sample for which the L -band value is higher than the IRAS 12 µm value. It has the coolest dust shell in the sample. It also has a very steep CLASS II spectrum, indicative of a very high extinction. IRAS 11159-5954 is the only object in the sample for which the peak of its SED in the near-infrared is higher than in the far infrared, showing that the grain emission is very weak.
5.5. Color-color diagrams
5.5.1. IRAS color-color diagram
In Fig. 1 we show the IRAS color-color diagram. The IRAS fluxes were converted to magnitudes according to the IRAS Explanatory Supplement (Beichman et al. 1984). The boxes defined by van der Veen & Habing (1988) are drawn in. According to this classification scheme, PNe are found in region V of the color-color diagram and AGB stars in region IV. In region VIII there may be some confusion from galaxies and young stellar objects, and in region IV an odd H II region may be present. The only object in region VIII is IRAS 15544-5332, which is not redshifted and shows Br in emission. The object is unresolved, and the Br emission is very weak. Hence it is unlikely to be an ultra-compact H II region, but we cannot completely rule out that it is an embedded young stellar object that is not hot enough to ionize its surroundings. IRAS 13416-6243 in region IV has Br in absorption and hence can be considered to be a post-AGB star. From the results in Fig. 1, it seems that in the IRAS color-color diagram no distinction can be made between post-AGB stars with Br in emission, absorption, or a flat Br spectrum. Post-AGB stars were expected to be located in a region in the IRAS color-color diagram between AGB stars and PNe (e.g. Volk & Kwok 1989; van der Veen et al. 1989; Hu et al. 1993). However van Hoof et al. (1997) found in their parameter study of the spectral evolution of post-AGB stars that they can follow a variety of paths in the IRAS color-color diagram. Consequently PNe and post-AGB stars can occupy the same region in the IRAS color-color diagram and the position in the IRAS color-color diagram alone cannot a priori give a unique determination of the evolutionary status of a post-AGB star. Our observations confirm this result.
The fact that our objects are mostly selected from the region where PNe were found and not in the region between AGB and PNe, may explain our high detection rate of Br emission (for comparison, the search by Käufl et al. (1993) for Br emission in a sample of 21 post-AGB stars resulted in only one detection).
5.5.2. Near-infrared color-color diagrams
In Fig. 2 and Fig. 3 we show the J -H versus H -K and the H -K versus K -L color-color diagrams, respectively. In the diagrams we notice that objects of CLASS II are redder than the objects of CLASS IV. The former are found in the top right part of the diagrams while the latter are located more towards the bottom left. The separation is most obvious in Fig. 3. For our sample, objects having K -L 1.5 mag are all of CLASS II and objects which have a maximum in the near-infrared beyond 2 µm have H -K 1.0 mag. Objects for which the stellar signature is more pronounced and which peak more towards shorter wavelengths in the near-infrared, are found more towards the bottom left in the diagrams.
To understand this effect, we would like to point out that for all but the coolest stars the intrinsic shape of the stellar continuum in the near-infrared can be approximated by a Rayleigh-Jeans tail. Since the shape of this tail does not depend on stellar temperature, the intrinsic infrared colors of these stars will not depend on stellar temperature either. Furthermore, since the colors of an A0V star are by definition zero, the infrared colors for most other stars are also close to zero, except when severe inter- or circum-stellar extinction is present. In this case the infrared colors will be positive and can be used as a crude measure for the extinction.
The mass loss rate during the superwind phase is very high. At the end of this phase the circumstellar dust will almost completely obscure the central star. During the post-AGB evolution, as the AGB shell expands, the shell will become more optically thin to stellar radiation. The stellar signature becomes more pronounced and will peak towards ever shorter wavelengths in the near-infrared. Consequently, it is expected that post-AGB stars will move from the upper right of the diagram to the lower left as the AGB shell expands and the circumstellar extinction becomes less. Generally speaking, we might expect that the objects in the top right of the diagram have left the AGB more recently than the objects in the lower left. However, we need to be cautious about such an interpretation because of the combined effects of many unknowns. The time it takes for the envelope to become optically thin will depend upon the mass loss rate at the tip of the AGB, the wind velocity, and the distribution of the mass in the circumstellar shell. If the AGB star had a non-spherical mass loss concentrated towards the equator, the central star could be visible along polar directions, while being completely obscured in equatorial directions. Hence, the observed amount of circumstellar extinction will depend on the viewing angle (Soker 1999). The interstellar extinction also affects the position of the objects in the color-color diagrams, as indicated by the arrow. The J -H versus H -K diagram is clearly the most affected by interstellar extinction. Eleven of the 16 objects are within one degree of the galactic plane, including the 6 objects in CLASS II. The two objects in CLASS IVb also have the highest latitude ( ). We calculated the extinction for our objects at a distance of 1 kpc and 4 kpc according to Hakkila et al. (1997): would be between 0.5 mag and 2.0 mag if the objects were at a distance of 1 kpc and between 2.1 mag and 6.2 mag with a median of 4.6 mag if they were at 4 kpc.
Objects with Br in emission, absorption, or a flat spectrum are well mixed in the diagrams. As discussed in the previous section it is unlikely that the stars with Br in emission have reached a temperature high enough to start to ionize the AGB shell and the emission probably originates in the stellar wind. Because we observe Br in emission from objects in CLASS II (e.g. IRAS 15144-5812), it seems that a fast stellar wind can be present at an early stage when the circumstellar shell is still very optically thick. We are currently trying to determine spectral types of our objects in order to determine their true post-AGB evolutionary status (Van de Steene et al., in preparation).
5.5.3. Combined near- and far-infrared color-color diagrams
In Fig. 4 and Fig. 5 we show the K -L versus - and K -L versus - diagrams, respectively. The K -L color roughly describes the evolution of the circumstellar extinction and is therefore a measure for the expansion of the AGB shell, while the - and - colors reflect the evolution of the spectrum of the circumstellar dust shell.
In Fig. 4 we see a weak trend towards cooler - colors with decreasing K -L color at first. When K -L 1.5 mag this trend seems to reverse, probably due to an increase in 12 µm flux. However, this trend will need to be confirmed with a larger sample.
In Fig. 5 we see a broad and weak trend towards hotter - colors with decreasing K -L color. Obviously the 25-µm flux increases faster than the 60-µm flux.
Theoretical calculations by Blöcker (1995) predict that directly after the star leaves the AGB, the stellar evolution is slow. This causes the AGB dust shell to cool when it expands. However, around K the evolution of the central star speeds up considerably and the grains in the AGB shell start to heat up again. This effect is most pronounced for silicate grains because, as the peak of the stellar spectrum moves into the UV, the efficiency with which these grains absorb light increases significantly. This causes the counter-clockwise loop which was first predicted by van Hoof et al. (1997) for oxygen-rich post-AGB stars in the IRAS color-color diagram. This effect could explain the reverse trend in Fig. 4, and also the trend in Fig. 5.
In Fig. 6 we plot the K -L versus L - diagram. The K -L color describes the evolution of the circumstellar extinction. The L - color relates the stellar component with the peak of the dust emission.
Because he K -L color is sensitive to the extinction and the L - color is insensitive to extinction, all CLASS II objects are found in the upper half of the diagram and CLASS IV objects in the lower half.
For our limited sample, all but one of the objects are found to the right of the dotted line. This is partly due to sample selection. As the arrow indicates, objects less obscured (e.g., due to orientation along the polar axis, or because of an optically thin shell, or less interstellar extinction), and especially the ones with cool central stars, could be situated below the dotted line.
IRAS 11159-5954 has a lower extinction than its L - color would indicate. This is an M-type star (Van de Steene et al., 2000, in preparation), which may still have ongoing mass loss. Its dust shell does not appear to be very thick and the star is very bright in the near infrared.
The objects to the right of the dashed line are extended in the near-infrared or the optical. IRAS 17009-4154 and IRAS 15553-5230 showed elliptical morphology. The former is very faint in the optical, the latter invisible. IRAS 13428-6232 shows a bipolar morphology in the near-infrared and and is also very faint in the optical. IRAS 16594-4656 is bright and was not observed to be extended in the near-infrared, though it showed a bipolar morphology in its HST image (Hrivnak et al. 1999). Probably they have a thicker cool circumstellar dust shell than objects located below the dashed line (more 25 µm-band flux) and/or their central star temperatures are higher (less L -band flux).
As the dust shell expands, the  magnitude will decrease a bit (for oxygen-rich post-AGB stars) or remain roughly constant (for carbon-rich post-AGB stars) (van Hoof et al. 1997). The L magnitude will increase, as the stellar temperature increases. The K magnitude will decrease as the star starts shining through the dust shell. Thus the L - values are expected to increase with decreasing K -L values as the shell expands and the star becomes hotter.
This is what we observe in Fig. 6, both for the extended and non-extended objects. A small L - color indicates that the system is young, hence the star is heavily obscured by the dust shell which is still close to the star. This is noticeable by the large K -L values. For objects with a larger L - color the AGB shell is already more detached and cooler. The extinction is less and this translates into a smaller value for the K -L color.
The K -L versus L - diagram (not shown) is very similar to the K -L versus L - diagram, but the spread is larger for K -L 1.0 mag, possibly because of an increase in 12-µm flux, as discussed in the previous section. Because of its relatively large 12-µm flux, IRAS 13416-6342 is located close to IRAS 17009-4154. It needs to be checked at higher resolution whether this source is extended.
In Fig. 7 we show the color-color diagram proposed by Ueta et al. (2000). Since the J -K color measures the circumstellar extinction, in a similar way as K -L color does, and because the evolution described above for the L - color is equally valid for the K - color, one would expect the same decreasing trend as in the previous diagram. However, extinction effects have a much stronger effect on the K - and J -K colors than on the L - and K -L colors, as can be seen from the arrows in both diagrams. Nearly all objects of CLASS IVa have J -K 2.5 mag. When the star becomes prominent in the near-infrared, the stellar maximum shifts through the K -band towards the J -band, causing a reversal in the evolutionary trend.
The extended objects are also found towards the right of the diagram. In this plot it is more obvious that IRAS 16594-4656 is the brightest of the 4 extended objects.
Ueta et al. (2000) had one M star, IRAS 04386+5722, offset towards blue K - color, similar to IRAS 11159-5954, but a bit bluer in J -K . Its position is indicated with the smallest grey box in Fig. 7.
The long-dashed line at J -K = 1.45 mag in Fig. 7 separates the regions what Ueta et al. call Star-Obvious Low level Elongated (SOLE ) and DUst Prominent Longitudinally EXtended (DUPLEX ) nebulae. One quarter of our objects would be classified as SOLE and three quarters as DUPLEX in this scheme. The grey regions show where the objects in their sample are located and are labeled with their acronyms. Our samples are obviously complementary: there is virtually no overlap! The difference may be caused in part by the selection criteria. They have selected known post-AGB candidates from the literature and imaged their nebulosities with WFPC2 in the optical. Consequently they chose optically bright post-AGB stars. We selected objects from the PNe region in the IRAS color-color diagram of which very few had an optical counterpart identified. Moreover, all our objects are located within 5 degrees of the galactic plane, while only 11 out of their 27 objects are (3 SOLE , 6 DUPLEX , and 2 stellar). The arrow in the diagram indicates a correction for = 5 mag of extinction. Larger interstellar extinction alone cannot explain why our samples appear different. If the objects in both samples are at similar distances, it is plausible that we have more massive central stars in our sample, and that they are evolving faster across the HR diagram. Further investigation is needed to understand the differences between both samples.
In summary, no distinction can be made between the objects showing Br in emission, absorption, or a flat spectrum, in any of the color-color diagrams. The trends we see in the near and far infrared are mainly due to the expansion, morphology, and dust properties in the circumstellar shell and the obscuration of the central star it causes. The trends show the expected evolution of the circumstellar shell. Whether the positions of the objects in the color-color diagrams can be directly related to the temperature and core mass of the central star needs further investigation.
© European Southern Observatory (ESO) 2000
Online publication: October 30, 2000